Engaging Youths for Accurate Risk Assessment in the Context of School Shootings

A series of high-profile school shootings, beginning in the 1990s, have focused the United States’ attention on the problem of school violence (1). Heavy news coverage and social media discussions about school shootings, such as the one in Parkland, Florida, in February of 2018, have stoked fears that the number of school shootings have increased dramatically in recent years. Indeed, “media contagion” may be a motivator for copycat shootings (2). On the other hand, many experts who study mass shootings, including those in schools, say that they are not happening more frequently but many of them are more deadly than past attacks. The numbers of school-associated violent deaths have not trended upward in the past 20 years (3). Additionally, the most recent federal data (4) show that other forms of student victimization are on a downward trend and that fewer students report fear of being hurt at school than in previous years. Children and youths remain much safer in schools than in the community. According to the National Center for Educational Statistics, “Between 1992–93 and 2015–2016, the percentage of youth homicides occurring at school each year remained at less than 3 percent of the total number of youth homicides. The percentage of youth suicides occurring at school each year remained at less than 1 percent of the total number of suicides” (3).

Extensive media coverage and the large number of deaths in high-profile school shootings have, understandably, generated a high level of public fear. This public fear “drove a dramatic shift in security-related policies and procedures in our nation’s schools. Many of those efforts proliferated in a desperate and well-intentioned effort to make schools safer, but they were often predicated on unrealistic appraisals of risk and misunderstanding about the nature of the actual threat” (1). Efforts such as zero tolerance—policies that seek to impose severe sanctions such as suspension and expulsion for minor offenses in the hopes of preventing more serious ones—have not demonstrated effectiveness. In fact, suspending and expelling children typically lead to worse overall outcomes for society in terms of healthy development and criminality (5). Measures such as arming teachers with guns also have yielded no evidence for effectiveness in decreasing school homicides (1).

Criminologists, media communications specialists, psychologists, and contagion-modeling mathematicians converge upon a similar conclusion: A key motivator of most mass shooters is the fame and power they perceive that they will achieve for their crimes. Many hope to garner as much or more media attention as the mass shooters who came before them and with whom they identify (6). A 2002 report of findings of the U.S. Secret Service Safe School Initiative (7) analyzed 37 school shootings and concluded that “there is no accurate or useful ‘profile’ of students who engaged in targeted school violence.” In the incidents analyzed, attackers were represented along the full social spectrum, from popular students to “loners.” The research found that the attacks were rarely, if ever, impulsive acts. In almost 80% of cases, the attackers communicated their ideas or plans before the incident. Typically, they told friends or other peer acquaintances—but rarely communicated a threat directly to the target of the attack. Few of these aware peers spoke up. “Kids get caught up in the code of silence that is so strong, they fail to see the larger picture” (8). In more than three-quarters of the incidents investigated, an adult had expressed concern about the attacker. “Many attackers felt bullied, persecuted, or injured by others prior to the attack” (p. 35), many had a history of suicide attempts or feelings of depression or desperation, and most had no history of criminal behavior (7). With this information, warning signs of violence may be identified and used to prevent serious premeditated violence.

Investigators have implemented automatic text-based analysis techniques to quantitatively and qualitatively analyze content of media messages sent by school shooters (8). Neuman and colleagues (9) identified a connection between social media and school shootings. Comparing texts written by six school shooters with 6,056 texts written by a comparison group of males using an automatic text-based analysis computer algorithm, they measured semantic similarities between words and texts. These semantic similarities (word vectors) represented four personality disorder dimensions: paranoid personality disorder, narcissistic personality disorder, schizotypal personality disorder, and depressivity. The words identified for text-based analysis were identified by Neuman et al. (9) based on the DSM-V criteria and Millon’s Personality Traits (10). They found that the shooters’ texts scored significantly higher on the Narcissistic Personality Disorder dimension, as well as the subcategories of the Humiliated and Revengeful dimensions (8, 9).